> ## Documentation Index
> Fetch the complete documentation index at: https://docs.argentos.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Providers

> Supported model providers — Anthropic, Ollama, MiniMax, Z.AI, and OpenRouter.

## Overview

ArgentOS supports multiple model providers, allowing you to mix local and cloud models across different tiers. Each provider has its own authentication, endpoint configuration, and model catalog.

<CardGroup cols={2}>
  <Card title="Anthropic" icon="a">
    Primary provider. Claude Haiku, Sonnet, and Opus.
  </Card>

  <Card title="Ollama" icon="server">
    Local open-source models. Zero cost.
  </Card>

  <Card title="MiniMax" icon="bolt">
    Alternative cloud provider with competitive pricing.
  </Card>

  <Card title="OpenRouter" icon="shuffle">
    Meta-provider routing to hundreds of models.
  </Card>
</CardGroup>

## Anthropic

The primary provider for ArgentOS. Supports Claude Haiku, Sonnet, and Opus.

* **Endpoint**: `https://api.anthropic.com/v1`
* **Auth**: API keys (`sk-ant-api...`) or setup tokens (`sk-ant-oat01-...`)
* **Models**: `claude-haiku-4-20250514`, `claude-sonnet-4-20250514`, `claude-opus-4-20250514`
* **Features**: Streaming, tool use, prompt caching (API keys only)

<Info>
  Prompt caching only works with standard API keys, not with Max subscription setup tokens.
</Info>

## Ollama (Local)

Run open-source models locally with zero cost.

* **Endpoint**: `http://localhost:11434`
* **Auth**: None required
* **Models**: Any model available in Ollama (Qwen3, Llama, Mistral, etc.)
* **Default**: `qwen3:30b-a3b` for the LOCAL tier

```bash theme={null}
# Install Ollama
curl -fsSL https://ollama.com/install.sh | sh

# Pull the default local model
ollama pull qwen3:30b-a3b
```

## MiniMax

An alternative cloud provider with competitive pricing.

* **Endpoint**: `https://api.minimax.io/v1` (portal API)
* **Auth**: Coding Plan key (`sk-cp-...`)
* **Models**: `MiniMax-M2.1` (general), `M2-her` (roleplay/character), `MiniMax-VL-01` (vision)

<Warning>
  MiniMax cannot handle mid-session fallback from Anthropic. Tool call history with Anthropic-format IDs (`toolu_01...`) causes HTTP 400 errors. MiniMax works best for fresh sessions or non-tool conversations.
</Warning>

## Z.AI

GLM-series models from Zhipu AI.

* **Endpoint**: Z.AI API
* **Auth**: Z.AI API key
* **Models**: `GLM-4.7`, `GLM-4.7-FlashX`, `GLM-4.7-Flash`

## OpenRouter

A meta-provider that routes to many model providers.

* **Endpoint**: `https://openrouter.ai/api/v1`
* **Auth**: OpenRouter API key
* **Models**: Access to hundreds of models from multiple providers

## Provider Configuration

Providers are configured through [auth profiles](/models/auth-profiles) in:

```
~/.argentos/agents/main/agent/auth-profiles.json
```

## Adding a Provider

```bash theme={null}
argent setup
```

<Tip>
  The setup wizard guides you through adding API keys for each provider. You can also manually edit the auth profiles file.
</Tip>